Experimental warming alters the community composition, diversity, and N2 fixation activity of peat moss (Sphagnum fallax) microbiomes.

Sphagnum-dominated peatlands comprise a globally important pool of soil carbon (C) and are vulnerable to climate change. While peat mosses of the genus Sphagnum are known to harbor diverse microbial communities that mediate C and nitrogen (N) cycling in peatlands, the effects of climate change on Sphagnum microbiome composition and functioning are largely unknown. We investigated the impacts of experimental whole-ecosystem warming on the Sphagnum moss microbiome, focusing on N2 fixing microorganisms (diazotrophs). To characterize the microbiome response to warming, we performed next-generation sequencing of small subunit (SSU) rRNA and nitrogenase (nifH) gene amplicons and quantified rates of N2 fixation activity in Sphagnum fallax individuals sampled from experimental enclosures over 2 years in a northern Minnesota, USA bog. The taxonomic diversity of overall microbial communities and diazotroph communities, as well as N2 fixation rates, decreased with warming (p < 0.05). Following warming, diazotrophs shifted from a mixed community of Nostocales (Cyanobacteria) and Rhizobiales (Alphaproteobacteria) to predominance of Nostocales. Microbiome community composition differed between years, with some diazotroph populations persisting while others declined in relative abundance in warmed plots in the second year. Our results demonstrate that warming substantially alters the community composition, diversity, and N2 fixation activity of peat moss microbiomes, which may ultimately impact host fitness, ecosystem productivity, and C storage potential in peatlands.

INTRODUCTION

High‐latitude peatlands account for a large proportion of soil C as permafrost and peat, which are susceptible to changing climatic conditions (Frolking et al., 2011; Gorham, 1991; Gorham, Lehman, Dyke, Clymo, & Janssens, 2012; Hayes et al., 2011; Limpens et al., 2008; Yu, Loisel, Brosseau, Beilman, & Hunt, 2010). A growing body of research has focused on the potential effects of climate on peatland plant dynamics (Camill, 1999; Dieleman, Branfireun, McLaughlin, & Lindo, 2015; Heijmans, Mauquoy, Geel, & Berendse, 2008; Mäkiranta et al., 2018), but little attention has been paid to the potential associated changes in plant microbiomes and the key ecosystem functions they perform. In general, plant microbiomes are susceptible to environmental perturbations such as drought (Naylor, Degraaf, Purdom, & Coleman‐Derr, 2017; Santos‐Medellín, Edwards, Liechty, Nguyen, & Sundaresan, 2017; Xu et al., 2018), nitrogen (N) deposition (Gschwendtner, Engel, Lueders, Buegger, & Schloter, 2016), and increased salinity (Yang, Hu, Long, Liu, & Rengel, 2016). Therefore, the plant microbiomes of peatlands are likely to respond to changing climatic conditions.

Peat mosses of genus Sphagnum dominate most northern peatland ecosystems, contributing up to 50% of aboveground production (Turetsky et al., 2012). Previous studies suggest that Sphagnum productivity and fitness are dependent in part on association of the plant host with a diverse microbiome (Bragina et al., 2014; Kostka et al., 2016; Warren et al., 2017), which also plays a direct role in C and N cycling. For example, diazotrophs are estimated to supply 40%–96% of N input to peatlands (Vile et al., 2014) with high accumulation of fixed N into plant biomass (Berg, Danielsson, & Svensson, 2013). Sphagnum‐associated methanotrophic bacteria can convert methane to CO2, which can then be utilized by the plant as a substrate for photosynthesis (Kip et al., 2010, 2011; Larmola et al., 2010; Raghoebarsing et al., 2005). Additionally, the Sphagnum microbiome may support the host in pathogen defense (Opelt et al., 2007; Shcherbakov et al., 2013) and stress tolerance (Opelt et al., 2007).

Despite the important role of the Sphagnum microbiome in moss productivity and nutrient cycling, the effect of warming on the Sphagnum microbiome is not well characterized. Sphagnum microbiomes are thought to be strongly impacted by abiotic factors (Bragina, Berg, Müller, Moser, & Berg, 2013; Leppänen, Rissanen, & Tiirola, 2015), and Sphagnum‐associated bacterial biomass was shown to change as temperature rises (Jassey et al., 2013). Therefore, warming is likely to influence microbiome structure, potentially altering plant and ecosystem function.

The objective of this study was to quantify the impacts of experimental warming on the abundance, diversity, community composition, and N2 fixation activity of the Sphagnum fallax microbiome over a 2‐year period. We hypothesized that warming would (a) decrease Sphagnum‐associated microbial diversity and (b) negatively impact rates of N2 fixation potential. To address our hypotheses, Sphagnum microbiomes were investigated by leveraging the Spruce and Peatland Responses Under Changing Environments (SPRUCE) experiment (Hanson et al., 2017), which consists of in situ whole‐ecosystem warming treatments from ambient temperature to ambient +9°C at the S1 bog of the Marcell Experimental Forest in northern Minnesota (Kolka, Sebestyen, Verry, & Brooks, 2011). The study focused on the diazotroph functional guild, which is believed to enable plant growth through N acquisition under the extremely nutrient‐limited conditions that characterize ombrotrophic bog ecosystems (Larmola et al., 2014; Limpens & Heijmans, 2008; Vile et al., 2014).

MATERIALS AND METHODS

Experimental site and warming experiment

The SPRUCE experiment located in the S1 bog of the Marcell Experimental Forest (47° 30.4760′ N; 93° 27.1620′ W; 418 m above mean sea level), Minnesota, USA, employs a regression‐based design at the whole‐ecosystem scale to produce nominal warming of ambient +0°C, +2.25°C, +4.5°C, +6.75°C, and +9°C in a Picea mariana–Sphagnum spp. raised bog ecosystem (Figure S1). Two enclosed replicate plots are employed for each warming treatment. The experimental site design was thoroughly described by Hanson et al. (2017). Briefly, the experiment consists of ten 12 m diameter plots with open‐top enclosures and two 12 m diameter plots without enclosures for ambient temperature experiments. The warming methodology combines air warming with deep‐peat heating from mild electrical resistance heaters to generate target warming levels superimposed over natural diurnal and seasonal variability (Hanson et al., 2017). Heating of the soil was initiated in June 2014 and atmospheric heating began in June 2015. Environmental data on humidity and relative humidity, surface moisture, and porewater pH are available for all years of the SPRUCE experiment (http://sprucedata.ornl.gov).

The S1 bog is an acidic and nutrient‐deficient ombrotrophic Sphagnum‐dominated peatland bog (surface pH ≤ 4, ~2 μM ammonium, <0.1 μM nitrate, and <0.1 μM phosphate; Kolka et al., 2011; Warren et al., 2017). Mean annual precipitation and air temperature are 768 mm and 3.3°C, respectively (Sebestyen et al., 2011). Porewater pH did not show differences across treatments (Griffiths & Sebestyen, 2016). Absolute humidity is maintained across treatments while relative humidity decreases with warming treatment (Hanson et al., 2017). Surface moisture measurements in the low‐density peat were not available in 2016 and 2017 due to equipment issues.

Sampling

To characterize the responses of the Sphagnum microbiome to warming, living green stem segments, 2.5 cm in length, were randomly collected from individual Sphagnum mosses in each plot. At the time of sampling, there were no visually apparent effects of stress on the Sphagnum plants. Plants were collected from hollows (i.e., depressed microtopographic positions) that were dominated by S. fallax with some S. angustifolium present. We targeted S. fallax for sampling but acknowledge that these species are difficult to distinguish from one another. In June 2016 (2 years after initiation of soil warming and a year after initiation of atmospheric warming), three individuals were collected for microbial community analysis and nine for acetylene reduction assay (ARA; from +0 and +9°C treatments only) from each experimental plot (total number of plots = 10). Following sampling in June 2016, elevated CO2 treatment was added to half the experimental enclosures; samples were only collected from ambient CO2 plots in 2017. Sampling efforts were limited to a minimal amount of plant material in order to avoid disturbing the plant community and interfering with long‐term efforts to measure ecosystem response to climate change. In June 2017, five S. fallax individuals were collected from each experimental plot (total number of plots = 5) for microbial community analysis: five for ARA and five for the 15N2 isotope incorporation assay (Table S1). After collection, samples were shipped to Oak Ridge National Laboratory on ice.

DNA extraction, PCR, and DNA sequencing

To characterize S. fallax microbiomes, each individual was pulverized to a powder in liquid N2, and DNA was extracted from 50 mg of powdered sample using the DNeasy PowerPlant Pro Kit (Qiagen) within 3 days of collection. The diversity and composition of S. fallax‐associated bacteria and archaea were determined using a high‐throughput sequencing of polymerase chain reaction (PCR) amplicons generated from the V4 region of small subunit (SSU) rRNA genes using primers 515F (5′‐GTGCCAGCMGCCGCGGTAA‐3′) and 806R (5′‐GGACTACHVGGGTWTCTAAT‐3′), as previously described (Kolton, Graber, Tsehansky, Elad, & Cytryn, 2017; Wilson et al., 2016). Diversity and composition of diazotrophic communities were assessed by targeting nifH, which encodes the nitrogenase reductase subunit and is a molecular marker for diazotrophs. Amplification of nifH with primers IGK3 (5′‐GCIWTHTAYGGIAARGGIGGIATHGGIAA‐3′) and DVV (5′‐TIGCRAAICCICCRCAIACIACRTC‐3′) yielded 396‐bp PCR products (Gaby & Buckley, 2014). The SSU rRNA and nifH amplicons were tagged with unique 10‐base barcodes (Fluidigm Corporation) and sequenced on the Illumina MiSeq 2000 platform according to standard protocols, including the use of negative control samples (Caporaso et al., 2012; Gaby et al., 2018; Gilbert et al., 2010). The 2016 amplicon libraries were sequenced at Georgia Institute of Technology, and the 2017 amplicon libraries were sequenced at the University of Tennessee, Knoxville. The SSU rRNA and nifH gene amplicon sequences have been deposited in the BioProject database (http://ncbi.nlm.nih.gov/bioproject) under accessions PRJNA407792 and PRJNA407800, respectively.

Sequence processing and analysis

Illumina‐generated SSU rRNA and nifH gene amplicon sequences were merged with PEAR (Zhang, Kobert, Flouri, & Stamatakis, 2014), and primers were trimmed with the Mothur v1.36.1 software (Schloss et al., 2009). The resultant sequences were quality filtered with Phred quality score cutoffs of Q30 and Q25 for SSU rRNA and nifH, respectively, using the standard QIIME 1.9.1 pipeline (Caporaso et al., 2010). Sequences of SSU rRNA and nifH genes were clustered into operational taxonomic units (OTUs) using the UCLUST algorithm with thresholds of 97% and 95% identity, respectively. Representative sequences were aligned using PyNAST (Caporaso et al., 2010) against the Greengenes core set (for SSU rRNA) or an alignment of nifH genes (DeSantis et al., 2006; Gaby & Buckley, 2014). Taxonomies of these high‐quality sequences were annotated with the Greengenes database (release 13_8) (DeSantis et al., 2006) or a manually curated nifH database (Gaby & Buckley, 2014) using the RDP classifier (Wang, Garrity, Tiedje, & Cole, 2007) with a minimum confidence threshold of 50%. SSU rRNA sequences classified as “chloroplast” (30% of reads) or “mitochondria” (<0.01% of reads) were removed to eliminate plant reads. An approximate maximum‐likelihood tree was constructed from the alignment of bacterial representative sequences using FastTree (Price, Dehal, & Arkin, 2009). Prior to diversity analyses, OTUs were rarefied to 3,500 and 1,000 reads per sample for SSU rRNA and nifH amplicons, respectively (Figure S2). OTU‐based alpha diversity was calculated based on Shannon diversity index (H′), which includes community richness and evenness in the metric. OTU‐based beta diversity indices were estimated based on Bray–Curtis distances.

Quantitative PCR amplification

Quantitative PCR assays were performed in triplicate using the StepOnePlus platform (Applied Biosystems, Foster City, CA) and PowerUp SYBR Green Master Mix (Applied Biosystems). Absolute quantification of SSU rRNA and nifH genes was conducted using primer pairs 331F (5′‐CCTACGGGAGGCAGCAGT‐3′)/518R (5′‐ATTACCGCGGCTGCTG‐3′) and PolF (5′‐TGCGAYCCSAARGCBGACTC‐3′)/PolR (5′‐ATSGCCATCATYTCRCCGGA‐3′), respectively (Muyzer & Waale, 1993; Poly, Monrozier, & Bally, 2001). Melting curve calculations and determination of T m values were performed using the polynomial algorithm function of the StepOnePlus Software (Applied Biosystems). Standard curves were obtained by serial dilution of standard plasmids containing target Escherichia coli K12 rRNA or Azotobacter vinelandii nifH gene fragments.

Acetylene reduction assay and 15N2 labeling

To determine the effect of warming on diazotroph activity, fresh S. fallax tissue from the ambient plots and the warming plots exposed to the highest temperatures (+9°C) in 2016, and all plots in 2017, were interrogated using the ARA as previously described (Warren et al., 2017). Briefly, for each condition, a 1.0–1.5 g sample of green, photosynthetically active or nonsenescent S. fallax was placed into a 35 ml glass serum bottle, which was stoppered with black butyl stoppers and sealed with an aluminum crimp seal; 10% headspace was replaced with 10% room air or 10% C2H2. Controls not amended with C2H2 did not produce detectable ethylene. All treatments were incubated for 1 week at 25°C with exposure to light. A gas chromatograph with a flame ionization detector (DRI Instruments, Torrance, CA) equipped with a HayeSep N column was used to quantify ethylene (C2H4). Accumulation of C2H4 was determined twice daily until C2H4 production was linear (~3 days). At the end of the incubation, samples were dried at 80°C for 48 hr to determine dry weight (for normalization of ARA rates) and Sphagnum tissue water content.

Incubations for 15N2 incorporation rates were set up as described above and supplemented with 10% headspace of 98% 15N2 (Cambridge Isotope Laboratories, Tewksbury, MA). After 3 days, samples were dried at 80°C and homogenized into a fine powder, and then dried samples containing ~50–100 µg N were analyzed for 15N by continuous‐flow isotope ratio mass spectrometry (Integra CN, SerCon, Crewe, UK). NIST standard 1575a was used to assess the accuracy and precision of natural abundance samples, whereas enriched standards of known 15N concentration (i.e., an (NH4)2SO4 solution ranging from 88 to 2,716‰ 15N) were used to ensure accuracy and precision of enriched samples.

Data analysis

Statistical analyses were conducted in R (R Core Team, 2015). To avoid pseudo‐replication (Hurlbert, 1984), microbial diversity and function values were averaged at the plot level to account for multiple individuals assessed from each plot. The relationship between warming and Shannon diversity index was quantified using a Pearson correlation, as previously employed with SPRUCE datasets (Richardson et al., 2018). Warming correlations with microbiome taxa were assessed with the Spearman correlation test, with p values corrected for multiple comparisons by the false discovery rate method. For all correlation tests, average June temperatures measured at soil surface were used for each respective year. A heatmap was generated from the relative abundances of diazotroph genera that exhibited significant differences (p < 0.05) and had >0.1% relative abundance in at least one treatment. The Mann–Whitney test was used to compare diversity between ambient plots with or without an enclosure structure. Beta diversity was visualized using nonmetric multidimensional scaling ordination (NMDS) based on Bray–Curtis similarity distances. A Mantel test with 999 permutations was used to determine temperature correlation with beta diversity across the treatments.

RESULTS

Effect of warming on the abundance, composition, and diversity of the microbiome

The 2016 amplicon library recovered 7,477 OTUs from 181,565 quality SSU rRNA sequences, and each sample yielded an average of 5,447 sequences after plant sequence removal. The 2017 amplicon library recovered 6,546 OTUs from 255,111 quality SSU rRNA sequences, and each sample yielded an average of 6,649 sequences after plant sequence removal. In both 2016 and 2017, S. fallax‐associated microbial communities were dominated by Proteobacteria (47%–62%) followed by Acidobacteria (14%–17%), with smaller contributions from candidate division WPS‐2, Cyanobacteria, Bacteroidetes, Verrucomicrobia, and Actinobacteria (<10% each; Figure 1a,b). Cyanobacteria was the only phylum to exhibit a statistical correlation with temperature, with a significant negative correlation both years (p < 0.05; Table 1). In both years, the diversity of S. fallax‐associated microbial communities decreased with warming (p < 0.05; Table 1 ; Figure 2a,c). S. fallax microbiome composition was structured by temperature treatment in 2016 and 2017 (Figure S3). Overall microbial abundance, as determined by qPCR, did not vary with warming treatment (r = −0.49, p = 0.1541; Figure S4). Water content of Sphagnum tissue was not correlated with temperature (Figure S5).

Figure 1

Effect of warming on overall microbial and diazotroph community composition in Sphagnum fallax microbiomes. Relative abundance of SSU rRNA or nifH gene sequences was determined at the phylum and family level, respectively, from triplicate samples collected in duplicate

Table 1

Sphagnum fallax microbial diversity, N2 fixation activity, and taxonomic groups' relationship with warming were measured using correlation tests

	2016	2017
Diversity
SSU rRNA gene	−0.680**	−0.960*
nifH gene	−0.790*	−0.920*
Function
15N2 incorporation	—	−0.930*
SSU rRNA Phyla
Cyanobacteria	−1.000*	−1.000*
Acidobacteria	0.800	0.900 +
nifH Families
Nostocaceae	0.900 +	0.700
Rivulariaceae	−0.900 +	−0.667
Chlorogloeopsidaceae	−1.000*	−0.707
Beijerinckiaceae	−1.000*	−0.700
Bradyrhizobiaceae	−0.900 +	−0.300

The relationship between warming and Shannon diversity was measured using the Pearson correlation. Warming correlations with microbial taxonomic group relative abundance measurements (SSU rRNA and nifH) were assessed using the Spearman correlation test, with p values corrected for multiple comparisons by the false‐discovery rate method. For all correlation tests, average June temperatures measured 0.5 m above soil were used for each respective year. Means of triplicate samples from each plot were used to test correlations with temperature (2016: total number of plots = 10, 2017: total number of plots = 5).

Significance is denoted as follows:

Only taxonomic groups with p < 0.05 for at least 1 year are displayed.

0.05 < p < 0.1;

p < 0.05;

p < 0.001.

Figure 2

Linear regressions (black lines) of Shannon diversity as a function of average measured temperature for each experimental plot. Shannon diversity was calculated for SSU rRNA gene sequences rarefied to 3,500 sequences per sample (a, b) and nifH gene sequences rarefied to 1,000 sequences per sample (c, d), Each data point represents an average of samples collected from each plot sampled in June 2016 (a, c) or 2017 (b, d). Pearson correlation coefficients (r) and p values are listed in the lower left of each panel. For all graphs, error bars indicate mean ± standard error

Response of diazotroph diversity, community composition, and N2 fixation activity to warming

A total of 1,523 OTUs were recovered from 21,192 high‐quality nifH sequences, and each sample yielded an average of 1,854 sequences for 2016. A total of 1,402 OTUs were recovered from 30,065 high‐quality nifH sequences, and each sample yielded an average of 1,371 sequences for 2017. Both years, nifH sequence libraries were dominated by Cyanobacteria (Nostocaceae, Chlorogloeopsidaceae, and Rivulariaceae; 60%–100%; Figure 1c,d). The relative abundance of Nostocaceae nifH amplicons increased with warming (up to 88% in the +9°C treatment in 2016), while Beijerinckiaceae and Chlorogloeopsidaceae declined with warming (Table 1; Figure 1c,d). Both years, the taxonomic diversity of diazotrophs in the S. fallax microbiome decreased with warming (p < 0.05, Figure 2b) and the diazotroph community was structured by temperature treatment (Figure S3). Diazotroph abundance, as determined by qPCR in 2016, did not vary in response to warming treatment (r = −0.45, p = 0.188; Figure S4).

The subset of diazotroph genera that showed significant differences between temperature treatments based on nifH sequence libraries is shown in Figure 3. Both years, Nostocaceae became dominant at warmer treatments (+4.5°C, +6.75°C, +9°C; Figure 3a,b). Rhodopseudomonas (Bradyrhizobiaceae) also increased with temperature in 2017 (Figure 3b). Methanotrophic genera (Methyloferula in 2016 and Methylocapsa and Methylobacterium in 2017) dominated nifH genera in low temperature (+0°C and +2.25°C) treatments. The relative abundance of the cyanobacterial genus Fischerella decreased with warming both years; the cyanobacterial genus Stigonema decreased with warming in 2016 only.

Figure 3

Heatmaps generated from the relative abundances of diazotroph genera that exhibited significant differences between temperature treatments (p < 0.05) and had relative abundance >0.1% in at least one treatment after nifH gene sequences were rarefied to 1,000 reads per sample for 2016 (a) and 2017 (b). Heatmaps were constructed for each year individually. The heatmap shows the Z‐score of the relative abundance of each genus within each sample with red indicating low relative abundance and red indicating high relative abundance. The nifH taxonomic affiliation is classified to the genus (left column) and family (right column) level

Rates of 15N2 incorporation were significantly negatively correlated with temperature, decreasing by ~50% in the +4.5°C, +6.75°C, and +9°C treatments relative to the +2.25°C and +0°C treatments (r = −0.93; p = 0.02; Figure 4a). ARA rates from 2016 decreased with temperature (r = −0.96, p = 0.04; Figure S6). With the exception of the +6.75°C treatment in 2017, ARA rates were ~45% lower in warming treatments, with a similar reduction across treatments (Figure S6), although ARA rates and temperature did not demonstrate a significant linear relationship with temperature (r = −0.47, p = 0.43; Figure S4). Nitrogen fixation rates measured by ARA and 15N2 incorporation were positively correlated with SSU rRNA and nifH gene diversity (r = 0.67, p = 0.04 and r = 0.77, p = 0.01; Figure 4b,c).

Figure 4

Linear regressions (black line) of 15N incorporation as a function of (a) average measured temperature treatment and (b) SSU rRNA gene and (c) nifH gene Shannon diversity values. Each point represents the average 15N incorporation of five replicates in a warming treatment. Assessment of 15N incorporation was measured for 2017 samples only. Pearson correlation coefficient (r) and p value are listed. Error bars indicate mean ± standard error

Experimental enclosure affect

We compared S. fallax microbial communities based on amplicon sequence diversity from +0°C treatments with and without an enclosure in 2016. The enclosure had no significant effect on SSU rRNA and nifH gene abundance (p = 0.33, p = 0.62), alpha diversity (p = 0.51, p = 0.58), beta diversity (16S rRNA: R 2 = 0.02, p = 0.4; nifH: R 2 = 0.01, p = 0.6), or community composition (Figure S7).

DISCUSSION

In this study, we found that two years of whole‐ecosystem warming led to a shift in the diazotroph community from a mixed community of Nostocales (Cyanobacteria) and Rhizobiales (Alphaproteobacteria) to a predominance of Nostocales. In general, N2 fixation activity and nifH diversity in the S1 bog were negatively correlated with temperature, consistent with a recent study (Kolton, Marks, Wilson, Chanton, & Kostka, 2019). Proteobacteria and Acidobacteria dominated SSU rRNA gene sequence libraries in all treatments, consistent with Sphagnum microbiomes in other bog systems (Bragina et al., 2012, 2014; Kostka et al., 2016; Opelt et al., 2007; Shcherbakov et al., 2013). Alphaproteobacterial diazotrophs (Rhizobiales, primarily Beijerinckiaceae and Bradyrhizobiaceae) were abundant at near‐ambient temperatures, consistent with a previous study of untreated indigenous diazotrophic peat communities at S1 bog (Warren et al., 2017). The high relative abundance of diazotrophic Rhodopseudomonas (Bradyrhizobiaceae) in warming plots in 2017 compared to 2016 suggests that this genus may be more resilient to multiyear temperature stress (Allison & Martiny, 2008). However, reduced diazotroph diversity and activity with increasing temperature suggests that the overall diazotrophic community was not resilient.

In general, warming was associated with decreased relative abundance of Cyanobacteria in the overall microbial community and increased relative abundance of specific diazotrophic cyanobacteria (Nostoc). This suggests that diazotrophic cyanobacteria (especially Nostoc) may be more resistant to warming than nondiazotrophic cyanobacteria. Reduced rates of N2 fixation activity at warmer temperatures were associated with a higher relative abundance of cyanobacteria, and this may be explained by the concept of “cheating” in certain Nostoc species. This phenomenon has been observed in the feather moss microbiome in which Nostoc are abundant, but express low levels of nifH and consequently may not fix substantial amounts of N for uptake by the host (Warshan et al., 2016). Warshan et al. (2016) observed that while Stigonema comprised <29% of the relative abundance of the cyanobacterial community, this group accounted for the majority of nifH expression. In our study, the branched heterocystous cyanobacteria Fischerella and Stigonema (Fay, 1992) decreased in relative abundance upon warming, becoming undetectable in the +9°C treatment, whereas Nostoc became more prevalent. This swap suggests that shifts within the cyanobacterial community composition may be responsible for the decreased rates of N2 fixation activity.

Changes in overall diazotroph community composition may also be responsible for differences in N2 fixation activity rates. Previous studies demonstrated that Alphaproteobacteria were the dominant diazotrophs in Sphagnum‐dominated peatlands (Larmola et al., 2014; Vile et al., 2014; Warren et al., 2017). In our study, we observed large reductions in the relative abundance of Alphaproteobacterial diazotrophs across warming treatments. While diazotroph abundance did not significantly decrease with temperature in 2016, nitrogen fixation rates declined in parallel with microbial diversity, suggesting that the change in N2 fixation was due to alterations in diazotroph community composition.

We acknowledge the possibility that the changes we observed in the S. fallax microbiome are not directly caused by warming, but rather by a secondary parameter that is controlled by temperature, such as soil moisture (Allison & Treseder, 2008), substrate availability (Biasi et al., 2008), or plant biomass (Rinnan, Stark, & Tolvanen, 2009; Zhang et al., 2016). For example, it is possible that the positive correlation we observed between Nostoc and temperature was actually due to fluctuations in soil moisture from temperature treatments across time, which in turn selected for desiccation‐tolerant Nostoc species (Sand‐Jensen, 2014). The response of S. fallax itself to temperature may also impact the microbial community through changes in C exchange or release of polyphenols and other secondary metabolites (Jassey, Chiapusio, Gilbert, Toussaint, & Binet, 2012; Verhoeven & Liefveld, 1997). Hence, the driving mechanisms causing decreased diazotroph diversity need to be further examined.

A reduction in diazotroph diversity and N2 fixation associated with warming may impact the ability of S. fallax to acquire N and subsequently production and fitness in a warmer climate. Reduced microbial diversity may make ecosystems more susceptible to environmental perturbations. In light of additional anticipated perturbations, such as N deposition or changes in precipitation patterns, these communities may be further impacted (Aanderud, Jones, Schoolmaster, Fierer, & Lennon, 2013). Indeed, a reduction in the richness and evenness of microbial communities in other ecosystems, such as soil or rhizosphere, is associated with a reduction in ecosystem functions, including nutrient cycling (Philippot et al., 2013; Wagg, Bender, Widmer, & Heijden, 2014), plant productivity (Bell, Newman, Silverman, Turner, & Lilley, 2005; Fierer et al., 2013; van der Heijden, Bardgett, & Straalen, 2008; Lau & Lennon, 2011), and plant resistance against pathogen attacks (Jousset, Schulz, Scheu, & Eisenhauer, 2011; Mendes et al., 2011). Additionally, Sphagnum mosses harbour potential latent plant pathogens; in many organisms, disease outbreaks are dependent both on the abundance of pathogens and the diversity of microbiomes (Bragina et al., 2011; Elad & Pertot, 2014; Tout et al., 2015). Alternatively, a reduction in diversity could correspond to a loss of pathogenic taxa, which might be beneficial to host plants. Nonetheless, pronounced temperature‐induced changes of S. fallax microbial communities may lead to functional changes in the microbiome.

Slight differences in diazotroph activity were observed between 15N incorporation and ARA methods across the warming treatments. This is consistent with recent peatland studies (Saiz, Sgouridis, Drijfhout, & Ullah, 2019; Warren et al., 2017) and is expected as acetylene is known to inhibit diazotrophic methanotrophs (Warren et al., 2017). Thus, 15N incorporation measurements are more appropriate for estimating N2 fixation in peatland ecosystems (Saiz et al., 2019). Nonetheless, both N2 fixation assays showed reduction of nitrogenase activity in warming treatments.

Climate warming was predicted to have a positive effect on Sphagnum diazotroph activity (Rousk, Pedersen, Dyrnum, & Michelsen, 2017) as conditions move closer to the temperature optimum of nitrogenase (25°C; Houlton, Wang, Vitousek, & Field, 2008; Vitousek, Hättenschwiler, Olander, & Allison, 2002). In contrast, we observed that warming had a negative effect on N2 fixation activity. In addition, a negative (Rousk, Sorensen, Lett, & Michelsen, 2015) or no (Gundale, Nilsson, Bansal, & Jäderlund, 2012) relationship between N2 fixation activity and temperature has been reported for boreal systems. This discrepancy between the observed and expected response of N2 fixation activity to temperature may be due to a number of mechanisms. First, differences in microbial community structure and composition may have a larger impact on rates of N2 fixation activity than environmental characteristics (Hsu & Buckley, 2009), especially if nitrogenase expression is dependent on community composition (Bellenger, Xu, Zhang, Morel, & Kraepiel, 2014; Darnajoux et al., 2017; Miller & Eady, 1988; Warren et al., 2017; Warshan et al., 2016). Second, Sphagnum mosses are photosynthetically responsive to temperature (Haraguchi & Yamada, 2011; Harley, Tenhunen, Murray, & Beyers, 1989), which can decrease photosynthate production and thus exudate availability to the microbiome. Third, elevated soil temperature has the potential to increase peat N mineralization (Bonan & Cleve, 1992; Rustad et al., 2001), and plant‐available NH4 + has been observed to increase in deep peat with warming in the SPRUCE experimental plots (C. M. Iversen, unpublished data). Elevated levels of available N can inhibit nitrogenase, resulting in decreased N2 fixation activity (Gundale, Deluca, & Nordin, 2011). Our findings show that changes in the S. fallax microbiome corresponded with decreased N2 fixation activity, suggesting a link between diazotroph community composition and N2 fixation activity. It remains to be determined whether peat N mineralization, microbe–microbe interactions, or host carbon allocation play a role in S. fallax diazotroph temperature interaction.

Furthermore, our results suggest that warming changes overall microbiome composition, which may ultimately decrease S. fallax primary production and/or the C storage capacity of peatland ecosystems. While Sphagnum growth was long considered to be nitrogen limited in Sphagnum‐dominated bogs (Aerts, Wallen, & Malmer, 1992; Damman, 1988; Vitt, 2006), recent evidence indicates that the plants may acquire sufficient nitrogen from diazotrophs (Berg et al., 2013; Larmola et al., 2014; Lindo, Nilsson, & Gundale, 2013; Novak et al., 2016; Vile et al., 2014). Thus, a reduction in the nitrogen acquired from the diazotroph community would likely alter Sphagnum production, fitness, and competitive advantage.

Our findings highlight the importance of incorporating the S. fallax microbiome into future investigations of warming in peatland ecosystems. The patterns we observed in temperature and diversity were consistent across years, highlighting the possibility that microbial diversity metrics could be incorporated into model predictions of S. fallax production. Some diazotroph families indicated potential resilience to warming, but additional studies are required to determine if resiliency is taxon specific or general to the diazotroph community. Global carbon model calculations for the IPCC RCP8.5 scenario predict that temperatures will increase an average of 2.6°C during summer and 4.8°C during winter over the next century (Riahi et al., 2011). Based on the results of our warming experiment, even a modest 2.5°C increase would decrease diversity and nitrogen fixation potential in the S. fallax microbiome. Intense warming events during summer are also anticipated (Riahi et al., 2011), and would likely further disrupt S. fallax microbiome diversity and diazotroph community structure.

CONCLUSIONS

Our results demonstrate that large‐scale above‐ and belowground warming was associated with a decline in the diversity of the overall microbial community along with diazotrophs of the Sphagnum microbiome. The observed reduction in microbial diversity correlated with a decrease in N2‐fixation activity. Warming was associated with a shift in the diazotroph community from a mix of Nostocales (Cyanobacteria) and Rhizobiales (Alphaproteobacteria) to a predominance of Nostocales. Moreover, N2 fixation activity and nifH diversity were negatively correlated with temperature across 2 years of whole‐ecosystem warming. Future efforts should extend investigations to the impacts of warming on the community composition and metabolic activity of microbiomes from other peatland plants, the role of Sphagnum host species in shaping the warming microbiome, and the potential for long‐term recovery or adaptation as resilience mechanisms.

Click here for additional data file.